jorisvandenbossche commented on pull request #11106: URL: https://github.com/apache/arrow/pull/11106#issuecomment-920061992
> Allows for writing code that would be replaced by something better if a UUID extension type is ever added Although you mention this as a "pro", you could also see this as a con, because *if* we at some point add a UUID extension type, then that would change the behaviour after this PR (while now we raise an error, and we would stop raising an error, which is a less problematic change in behaviour). > The issue could also almost be solved by a pandas UUID extension type except pandas read_sql_query supports dtype but not a type-mapping mechanism that would allow a pandas UUID extension type to automatically be used if the column type was UUID. I have not been really involved in the pandas read_sql code the last years, but that sounds as something that could be improved there (allowing to specify a (extension) dtype per sql type). Pandas already has such an internal mapping to fix up dtypes when reading from a table (https://github.com/pandas-dev/pandas/blob/5872bfe1c713ddb27b337e5b6549bf497e44834b/pandas/io/sql.py#L1191-L1218), but no way to extend this. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: github-unsubscr...@arrow.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
